7 Replies Latest reply on Oct 18, 2012 1:14 AM by jcnars

    ZFS SA share as FRA on a Exadata rack

      Looking to see if ZFS Storage Appliance (we are buying one . period.) can be used as the FRA for the Exadata instead of the regular +RECO DG.
      Folks out there went thru' a similar thought-process? I've read that ODA can be connected to a ZFS SA for HCC & backup purposes, but wondering if there's some official document/Metalink note that says, ok Exadata can have FRA on ZFS.

      Obviously, **if** we have our FRA on ZFS, it just takes a simple RMAN 'switch to copy' and avoids the pain in the neck of tape speeds.
        • 1. Re: ZFS SA share as FRA on a Exadata rack
          I don't know of any official documentation that says anything about using a ZFS for the FRA, but keep in mind that it won't be within an ASM diskgroup, but would be mounted via NFS.

          While the "switch to copy" idea is nice, in practice it would most likely be very painful because you wouldn't be able to get any smart scans off of the ZFSSA.
          • 2. Re: ZFS SA share as FRA on a Exadata rack
            ZFS is an NFS mount and behaves no differently from any other NFS mount for purposes of locating the FRA.

            I've chosen to leave the FRA where it is but archive my redo logs across the IB 40G connection as well as use the ZFS to hold backups.

            ZFS does have HCC, lots of value there, but it does not have other Exadata features upon which you may be dependent for running a standby.

            My ZFS page is here:
            and I am adding to it on a regular basis as I accumulate additional screen captures in my spare time.
            • 3. Re: ZFS SA share as FRA on a Exadata rack
              Marc Fielding
              Keep in mind that putting your FRA on ZFSSA also involves archivelogs, making the ZFSSA a point of failure for your database. It could potentially make you more vulnerable than you are now.

              If you're looking for protection against a double disk failure, consider ASM high redundancy or a data guard standby.
              • 4. Re: ZFS SA share as FRA on a Exadata rack
                Thank you gentlemen.

                @Andy, just to make sure we're on the same page, All my datafiles will be in +DATA DG benefiting from Exadata optimizations. I take your answer to mean that my 'switch to copy' won't benefit from the RMAN non-smart scan optimization that Enkitec gurus talk in their book (pages 44 & 295).

                Simply superb. Much better than my ZFS simulator pages !

                Thanks. Yes, I am aware of that...that's the only "online" piece that'll touch the FRA, so I've thought about retaining the existing +RECO DG just for the archive log dest and housing the backup in ZFS shares.

                Still wondering if there's a godly doc/link/MAA wp from Oracle that says go ahead & put your Exadata FRA on ZFS.
                Or, even a link that says, DON'T DO IT
                • 5. Re: ZFS SA share as FRA on a Exadata rack
                  Marc Fielding
                  In this type of config, you may want to set two archivelog destinations: one on your +RECO diskgroup, and one on your ZFSSA.                                                                                                                                                                                                                                                       
                  • 6. Re: ZFS SA share as FRA on a Exadata rack
                    "The Oracle Database Fast Recovery Area should not be put on the Sun ZFS Storage Appliance,
                    and should remain on the Exadata Storage Servers that are part of the Exadata Database Machine"

                    See page 19 of http://www.oracle.com/technetwork/database/features/availability/maa-wp-dbm-zfs-backup-1593252.pdf
                    1 person found this helpful
                    • 7. Re: ZFS SA share as FRA on a Exadata rack
                      That nails it. An idea whose time will never come.
                      Thanks much Dan.